课本page88下3.11公式的证明
-
已知原图像灰度值r的概率密度函数 p r ( r ) p_{r}(r) pr(r),变换函数 s = T ( r ) s=T(r) s=T(r) 且严格单调递增,一阶导数存在,则变换后的s的概率密度函数(PDF)为: p s ( s ) = p r ( r ) ∣ d r d s ∣ p_{s}(s) = p_{r}(r) |\frac{dr}{ds}| ps(s)=pr(r)∣dsdr∣ 或写成 p s ( s ) = p r ( h ( s ) ) ∣ h ′ ( s ) ∣ p_{s}(s) = p_{r}(h(s)) |h^{'}(s)| ps(s)=pr(h(s))∣h′(s)∣ 其中 h ( s ) = T − 1 ( s ) = r h(s) = T^{-1}(s) = r h(s)=T−1(s)=r
-
证明
- F S ( s ) = P ( S ≤ s ) = P ( T ( r ) ≤ s ) F_{S}(s) = P(S \le s) = P(T(r) \le s) FS(s)=P(S≤s)=P(T(r)≤s)
- F S ( s ) = P ( r ≤ h ( s ) ) = F R ( h ( s ) ) F_{S}(s) = P(r \le h(s)) = F_{R}(h(s)) FS(s)=P(r≤h(s))=FR(h(s))
- p s ( s ) = p r ( h ( s ) ) ⋅ h ′ ( s ) p_{s}(s) = p_{r}(h(s)) \cdot h^{'}(s) ps(s)=pr(h(s))⋅h′(s)
-
由于是严格单调递增,所以一阶导不加绝对值没关系